Cancer is among the top ten causes of death in the world but in spite of the efforts of the pharmaceutical companies and many governmental organizations, new and more effective drugs are urgently needed. Computer-assisted studies have been widely used to predict anticancer activity taking into account different molecular descriptors, statistical techniques, cell lines, and datasets of congeneric and non-congeneric compounds. This paper describes a QSAR study and the successful application of 3D-MoRSE descriptor for developing Linear Discriminant Analysis (LDA) to predict the anticancer potential of a diverse set of indolocarbazoles derivatives. Despite the structural complexity of this sort of compounds the used variables are able to identify the most remarkable features like the incidence of polarizability of the substituents and the interatomic distance in the 7-azaindole moiety in the antiproliferative activity. A comparison with other approaches such as the Getaway, Randić molecular profile, Geometrical, RDF descriptors, was carried out showing the model with 3D-MoRSE descriptor resulting in the best accuracy and predictive capability. An LDA-based desirability analysis was conducted to select the levels of the predictor variables, in other words, the values of the independent variables which should generate more desirable anticancer chemicals, i.e., with higher posterior probability to be classified cytotoxic.